TL;DR
This study presents the first comprehensive phenome-wide heritability analysis using the UK Biobank, estimating the contribution of common genetic variants to 551 diverse human traits. It reveals that heritability is not fixed but can be significantly influenced by demographic factors like age, sex, and socioeconomic status, highlighting the importance of population-specific considerations in genetic research and risk prediction.
What Problem Does This Paper Address?
Understanding the heritability of human traits – the proportion of phenotypic variation attributable to genetic factors – is crucial for fields ranging from evolutionary biology to personalized medicine. While twin studies have long established that most complex traits are heritable, estimating the precise contribution of common genetic variants (SNP heritability) has been a challenge. Furthermore, heritability estimates can vary significantly between populations due to differences in genetic background and environmental influences. This study tackles these issues by:
- Providing a comprehensive phenome-wide heritability estimate: Moving beyond single traits, it aims to quantify the genetic contribution across a broad spectrum of human phenotypes.
- Leveraging a large, well-characterized dataset: The UK Biobank offers an unprecedented opportunity to study heritability in a large, population-based cohort.
- Investigating the influence of demographic factors: It addresses the critical question of whether heritability estimates are universal or if they vary based on age, sex, and socioeconomic status, which could have implications for interpreting genetic findings and developing predictive models.
Methods
The researchers employed a computationally efficient, moment-matching approach to estimate SNP heritability. This method is equivalent to Haseman-Elston regression and phenotype-correlation genetic-correlation regression, offering a lower computational and memory demand compared to traditional ReML algorithms, making it suitable for large datasets like the UK Biobank.
Key methodological aspects include:
- SNP Heritability Estimation: They calculated the proportion of phenotypic variance explained by common genetic variants. This provides a lower bound for narrow-sense heritability, as it doesn't capture non-additive genetic effects or rare variants.
- Covariate Adjustment: To account for potential confounding factors, analyses included covariates such as genotyping array, UK Biobank assessment center, age at recruitment, and the top 10 principal components of the genotype data. Additional covariates like sex and handedness were included where appropriate.
- Stratification Analyses: To examine the moderating effects of demographic variables, heritability was estimated separately for different age groups (using a sliding window approach), sexes, and socioeconomic status (SES) levels (quantified by the Townsend deprivation index).
- Statistical Rigor: Multiple testing was accounted for by calculating the effective number of independent traits and applying corrections to p-values.
Datasets
The primary dataset used in this study is the UK Biobank. This is a large-scale, population-based cohort study that enrolled approximately 500,000 participants aged 40-69 years between 2006 and 2010. The study collected extensive phenotypic data through questionnaires, physical and biological measurements, and electronic health records, alongside genome-wide genotype data.
- Genotype Data: An interim release of genotype data comprised 152,736 samples, genotyped using Affymetrix arrays (UK BİLEVE Axiom and UK Biobank Axiom arrays), covering approximately 800,000 markers.
- Phenotypic Data: The study analyzed 551 complex traits, including quantitative phenotypes and disease codes. These traits were categorized into 11 phenotypic domains.
The paper does not introduce new benchmark datasets.
Key Insights and Ideas
- Heritability is Not Universal: The most striking insight is that heritability is not a fixed property of a trait but is demonstrably influenced by population characteristics. This challenges the notion of a single heritability value for any given trait and emphasizes the need for context.
- Demographic Factors Matter: Age, sex, and SES are not just background variables; they can significantly moderate the genetic contribution to various phenotypes. This suggests that genetic effects might be amplified or diminished in specific demographic subgroups.
- Comprehensive Phenome-Wide Approach: By analyzing 551 traits, the study provides a broad overview of the genetic architecture of human variation, revealing that common genetic variants contribute to a wide array of traits and diseases.
- Methodological Advancement: The use of a computationally efficient moment-matching method enabled the analysis of such a large number of traits and participants, pushing the boundaries of what's feasible in heritability studies.
Key Results
- Broad Heritability: SNP heritability estimates were obtained for 551 traits, with common genetic variants showing an influence on most investigated traits.
- Highly Heritable Traits: Traits like human height (0.685+/-0.004), skin color (0.556+/-0.008), and rheumatoid arthritis (0.821+/-0.046) exhibited high SNP heritability.
- Environmentally Driven Traits: Conversely, traits such as duration of walks, frequency of stair climbing, and pain in joints showed near-zero heritability, indicating a stronger environmental influence.
- Sex Differences: Significant sex differences in heritability were observed for several traits, with women generally showing higher heritability for traits like waist circumference and blood pressure.
- Age Effects: A majority of physical measures showed a decreasing trend in heritability with age, suggesting that genetic influences may wane over the lifespan, possibly due to cumulative environmental effects.
- SES Moderation: Education was the only trait where SES showed a significant moderating effect, with heritability increasing with higher SES. This finding indirectly supports earlier studies on the SES moderation of IQ heritability.
Limitations and Future Work
The authors acknowledge several limitations:
- Phenotype Availability: The analysis was limited to traits with sufficient sample size for precise heritability estimates, meaning low-prevalence diseases (e.g., psychiatric disorders) were underrepresented.
- Population Prevalence Assumption: For binary traits, they assumed population prevalence equals sample prevalence, which may not hold for under-ascertained diseases.
- Self-Reported Data: A substantial fraction of phenotypes relied on self-report or diagnostic codes, which can be noisy and have varying specificity.
- Binarization of Variables: Categorical variables were binarized for analysis, which might not optimally represent their variation.
- Generalizability: Heritability is population-specific, so estimates may not generalize to other settings or ancestry groups.
- Genetic Architecture Assumptions: The methods used assume causal SNPs are randomly spread and have Gaussian effect sizes, which may not always be true.
Future work could explore:
- More Sophisticated Models: Investigating methods that account for MAF properties and region-specific LD heterogeneity of causal variants.
- Imputed Data: Estimating heritability using imputed data to capture rare variants and improve heritability estimates.
- Deeper Investigation of Modifying Factors: Further exploring the genetic architecture and finding the most appropriate models for individual traits, especially for those where demographic factors show significant moderation.
- Broader Phenotype Inclusion: Developing methods to include low-prevalence traits and improve the validity of phenotypes derived from electronic health records.
Summary updated